深度学习理论及其在语音识别领域的应用  

Deep Learning Theory and Its Application in Speech Recognition

作　　者：杨俊安[1,2] 王一[1,2] 刘辉[1,2] 李晋徽[1,2] 陆俊[1,2] 

机构地区：[1]解放军电子工程学院,安徽合肥230037 [2]安徽省电子制约技术重点实验室,安徽合肥230037

出　　处：《通信对抗》2014年第3期1-5,共5页Communication Countermeasures

摘　　要：深度学习是模式识别和机器学习领域最新的研究成果,它以强大的建模和表征能力在图像和语音处理等领域取得了很好的应用。将深度学习引入到电子对抗领域的语音识别中,首先简单介绍了深度学习的基本理论,随后阐述了目前语音识别领域中语种识别、说话人识别和关键词检出存在的问题,并利用深度学习方法对这些突出的问题加以解决。Deep Learning is an emerging area of pattern recognition and machine learning. It has been successfully used in image and speech processing by its more powerful modeling and representational abilities. In this paper, as an attempt to share this expertise with the researchers in the area of electronic warfare, we firstly discuss the basic principles of deep learning, and then we provide a survey on the existing language recognition, speaker identification and keywords spotting technologies. Finally, in order to overcome the inherent flaws in these technologies, we use deep belief network as examples to improve their performances, experimental results show that with the help of deep learning, we can achieve better speech recognition results than ever before.

关 键 词：深度学习 深度信念网络 语音识别 特征提取 声学建模 

分 类 号：TN912.34[电子电信—通信与信息系统]